A second attempt
Agents who have a productivity larger than \(50\) are predicted to feel treated kindly, if they are monitored by their principals. Agents with lower productivity should feel treated unkindly. If agents are reciprocal, they are predicted to pass back this perceived kindness (unkindness) by performing better (worse) than in the first stage.
Practically, one has to take into consideration that (1) it is easier for productive agents to diminish their performance, as they have a higher starting point (and therefore, more room for deterioration) and that (2) it is harder to pass back kindness (by being more productive) than to pass back unkindness (by shirking).
| Pre-Registerd | RDD | Post-Collection | |||||||
|---|---|---|---|---|---|---|---|---|---|
| Predictors | Estimates | CI | p | Estimates | CI | p | Estimates | CI | p |
| (Intercept) | -7.33 | -15.07 – 0.40 | 0.063 | 2.62 | -17.15 – 22.38 | 0.793 | -2.71 | -5.27 – -0.14 | 0.039 |
| productivity | 0.14 | -0.02 – 0.30 | 0.096 | -0.14 | -0.67 – 0.38 | 0.590 | |||
| productiveTRUE | 3.29 | -43.28 – 49.86 | 0.888 | 4.19 | 0.17 – 8.22 | 0.041 | |||
|
productivity * productiveTRUE |
0.07 | -0.80 – 0.94 | 0.873 | ||||||
| Observations | 79 | 79 | 79 | ||||||
| R2 / R2 adjusted | 0.036 / 0.023 | 0.057 / 0.019 | 0.053 / 0.041 | ||||||
\[ \begin{align} U_i(l_i(h),(b_{ij}(h))_{i\neq j},(c_{iji}(h))_{i\neq j}) & = \pi_i(l_i(h),b_{ij}(h))_{i\neq j}) \\ & + Y_{ij} \cdot \kappa_{ij}(l_i(h), b_{ij}(h))_{i\neq j}) \cdot \lambda_{iji}((b_{ij}(h))_{i\neq j},(c_{iji}(h))_{i\neq j}) \\ & - c_i(l_i(h)) \\ & + \sigma_i(l_i(h)) \end{align} \]
Focusing on agents who are not being ignored by their principals, one cannot compare the first stage’s productivity with the second stage’s performance, as the strategic environment is completely different. After all agents that are paid randomly have no incentive to exert any effort at all.